An experimental examination of fisheries with concurrent common pool and individual quota management.

Anderson, Christopher M. ; Uchida, Hirotsugu

I. INTRODUCTION

How user communities exploit and develop institutions to manage common pool resources attracts attention from across the social sciences (e.g., National Research Council 2002). Dynamic common pool resources such as aquifers, forest soils, and fisheries are a focus because of their potential to sustain income, and provide food and water security, particularly in traditional communities, and in vulnerable communities in the developing world (e.g., Arnason, Kelleher, and Willman 2009; Baland and Platteau 1996). Economic analyses have been directed primarily at identifying extraction levels that central planners can impose in order to maximize the highest sustainable flow of resource benefits (Conrad 2010; Wilen 2000), while political scientists, sociologists, and anthropologists have studied the governance structures that establish use norms and policies, and the conditions under which users can successfully manage the resource without central intervention (e.g., Agrawal 2002; Ostrom 2009).

Recent trends in management have introduced economically significant examples of dynamic resources where a central planner constrains the full extent of the interseason externality, ensuring sustainable resource levels and extractions, but other externalities persist within seasons that have the potential to dissipate much of the rent. The most prevalent examples are in fisheries, where management agencies use biological fish stock assessments to establish a hard seasonal total allowable catch (TAC) for each fishery to ensure stock health. In many fisheries, "catch share" programs allocate shares of the TAC as quota to groups, or to individuals who then can or must form cooperatives, to manage intraseason externalities caused by concurrent production congestion, in-season growth patterns, or market timing. Interest is added because these groups, called sectors in many fisheries, are often self-identifying, and multiple sectors participate in each fishery; harvesters can affiliate around harvesting methods, management preferences, geography, or any other factor. On the one hand, this strategy could be a method for carving users into groups that meet Ostrom's (2009) conditions and thus supporting self-management of intraseason externalities (e.g., Cunningham and Bostock 2005; Aoki and Hayami 2001; Townsend, Shotton, and Uchida 2008; Wilson, Nielsen, and Dengbol 2003); on the other hand, it could be a technique for breaking political gridlock preventing more efficient economic management of the inter- or intraseason externalities (e.g., Acheson and Gardner 2011; Kuperan and Sutinen 1998).

While such sectorization programs are thought to have many benefits such as an increase in legitimacy, incorporating local knowledge, and getting managers out of the business of setting individual allocations, prospective and retrospective economic analysis of these policies is difficult because, if different sectors choose different regulations, the fishery can be regulated by multiple management systems concurrently. This presents two significant challenges to analysts assessing the effects of such catch share programs. First, they must figure out not only how harvesters in each sector will respond to the harvesting strategies of others in their sector but also how the sectors will strategize against one another. Where contemporaneous congestion or price externalities are present, some sectors' rules might permit members to strategically adjust their harvesting behavior to respond to harvesting strategies of other sectors. Thus, a theoretical and empirical framework requires harvesters to make daily decisions that are subject to cumulative annual constraints within a dynamic game, with different types of players reflecting each sector. In addition, the counterfactual policy environment against which the sector program will be assessed must take into account the shifts in targeting and market timing behavior that not only reflect changes in the rules that apply to sector members but also the different rules under which harvesters are operating.

A second analytical challenge is that, since all harvesters hold a license to the same fishery regardless of the sector they join, harvesters can typically change sectors or whole sectors can change how they manage themselves. Thus, a forward-looking analysis of the sector program needs to anticipate how the rules and practices in the fishery might evolve over time, as harvesters gain experience under one set of rules and observe outcomes under other sets of rules; those in less successful sectors might migrate to sectors that are more successful. To the extent that success is determined by the management rules each sector selects for its catch allocation, this may prove to be an evolutionary selection process for management rules.

Understanding how differently managed groups pursuing the same stock will react is not entirely new in fisheries. Neighboring nations will often manage transboundary stocks slightly differently, but these situations typically result in TAC-setting strategies that reflect a common pool resource management problem between governments (e.g., Grand Banks fish wars, Levhari and Mirman 1980; northeast Atlantic mackerel, Hanneson 2013 and Hotvedt 2011). Similarly, recreational anglers frequently find themselves differently managed than commercial harvesters, but the differences are driven by differences in objectives and production functions, not designed to encourage one group to respond to others.

What distinguishes our cases of interest is that a single management framework allocates shares of a single TAC among groups of users to allow different harvesters to pursue different management. This creates a situation where at least one of the groups can choose management that gives itself flexibility in pursuing their share in strategic response to the others. An early example arose under the Pacific Halibut Commission, where the jurisdictional boundary between Canada and the United States allowed flexibility for each country to select separate rules for harvesting their share of the Commission-set TAC. British Columbia (BC) established an individual transferable quota (ITQ) system that allowed harvesters to suspend landings during the Alaskan derby, when prices were low (Casey et al. 1995); importantly, Alaska later adopted an ITQ scheme, changing the value of the Canadian ITQ system to BC harvesters. A current similar example arises from the Atlantic States Marine Fisheries Commission's management of striped bass by allocating portions of the coastwide TAC to states. In Virginia, striped bass harvesters developed an ITQ system in order to extend the season, which allows them to avoid market gluts caused during neighboring North Carolina's few-day derby fishery (Crosson 2011).

Increasingly, however, single management bodies are incorporating some form of sector formation within the catch share frameworks within a single jurisdiction. For example, the Chignik Coop among salmon harvesters in the Aleutian Islands of Alaska saw 77 of 99 harvesters elect to join a sector designed to minimize harvesting costs by limiting the number of members who fished while sharing revenue, while the rest chose to remain in a common pool derby fishery (Deacon, Parker, and Costello 2008; Knapp 2008). In Rhode Island, the eight vessels of the Rhode Island Fluke Conservation Cooperative (Scheld, Anderson, and Uchida 2012) accepted a hard TAC in exchange for a quota allocation of summer flounder. (1) They elected to harvest around half of their quota allocation when the common pool fluke fishery--managed as seasonal common pool derbies--was closed, and sector members were the only harvesters allowed to land fluke. Nearby, the Cape Cod Hook sector (Pinto da Silva and Kitts 2006; Verani 2007) and the Cape Cod Fixed Gear Sector (Georges Bank Cod Fixed Gear Sector 2010) hold collective allocations of Northeast Multispecies fish that they can harvest independent of many of the regulations that dominate the rest of the fishery. In each of these cases, the sectors were established in order to exploit opportunities for adjusting harvest timing from that required by the governing management plans in order to receive better prices.

On the basis of the experience with the Cape Cod sectors, and the small concurrent effort in Rhode Island, the entire Northeast Multi-species Fishery adopted sector management in 2010, leading to the establishment of 17 different self-identifying sectors that received collective quota allocations for the included stocks to manage in any way they wish, and a common pool that fishes competitively for the remaining quota associated with their history. While sectors formed primarily along gear, target species, and geographical lines, they were able to adjust the timing of their participation in the groundfish fishery to increase the landings and revenue from groundfish and non-groundfish species each vessel pursues under different management plans (Kitts et al. 2011). To begin to develop and test frameworks for understanding what happens with single fisheries under multiple management systems, we develop a controlled economic experiment to demonstrate the strategic types of harvest timing changes that occur in a fishery with two management systems: one sector that operates as a common pool (CP), and one that operates using individual quotas (IQ)--a common implementation of an efficiently managed sector with strong property rights. (2) In a novel quasi-continuous fishery game with a contemporaneous price externality, subjects individually choose how much effort to exert in each week of a 52-week fishing season, knowing that prices will be lower in weeks when total landings are higher. After subjects gain experience in a CP fishery, an IQ fishery, and a fishery where half the subjects are CP managed and half are IQ managed, they are given the opportunity to choose which management system they prefer, and fish for several seasons under their chosen management.

The next section presents the model that is the basis for our experiment. Then, the experimental environment and treatments are explained. Section IV presents the results, and Section V concludes with an emphasis on how harvesting strategies in multiply managed common pool resources can be importantly different from those where all harvesters fall under the same set of management incentives.

II. THE FISHERY ENVIRONMENT

Game theoretic models of common pool resources that have been widely studied are predominantly one-shot, static games in which players choose a level of extraction or investment effort (e.g., Walker, Gardner, and Ostrom 1990). Aggregate payoffs are inversely parabolic in total effort, but increasing private payoffs conditioned on a level of total effort lead to predicted Nash equilibrium extraction levels that are well above the social optimum. In experimental tests, Nash predictions are broadly consistent with the data in formulations capturing cost-based congestion externalities motivated by groundwater pumping (Gardner, Moore, and Walker 1997) and other production externalities; in cases with dynamic externalities, the observed outcomes can be even less efficient than the Nash prediction (Walker and Gardner 1992).

These games are not well suited to study seasonal quotas in fisheries because they emphasize the total levels of extraction within a harvest period. In a fishery context, this corresponds to the problem of setting the annual TAC, or harvesters simultaneously choosing their annual landings. This abstracts from the incentives that are most directly affected by the rules of catch share systems: the distribution of effort across individuals (allocations) and the distribution of effort across time. Thus, we use a common pool resource model wherein agents choose their level of fishing effort in a number of periods during a fishing season, subject to constraints on total harvest. This model is similar to Fell (2009), but has fewer dynamic processes to emphasize the market externality, simplify explanation to experimental subjects, and facilitate closed form solutions.

Consider a common pool resource which can be extracted through a season of t = 1, ..., T periods by i = 1, ..., I users. In each period, each user chooses an effort [e.sub.it], leading to total effort [E.sub.t] = [[SIGMA].sub.i] [e.sub.it], and the resulting total harvest H([E.sub.t]). Then, the payoff in period t to a user choosing [e.sub.it] is given by:

(1) [pi]([e.sub.it], [E.sub.t]) = [a - bH([E.sub.t])] x [([e.sub.it]/[E.sub.t]) H([E.sub.t])] - [gamma][e.sub.it].

The first term in square brackets is the price, with a linear demand curve that slopes down proportional to the total harvest at time t if b > 0. The second bracketed term is i's share of the total harvest, proportional to her share of the total effort. A production externality can be imposed by specifying a concave function for H(*). Finally, the rightmost term captures the linear cost associated with the chosen effort level. In this article, we specify a linear H(*) and focus on a price externality by setting b > 0. (3)

User i will choose a vector of [e.sub.it] that will be an individual profit-maximizing best response to the [e.sub.jt] choices of other users, subject to the constraints imposed by management. This article tests hypotheses arising from the game theoretic model when one group of harvesters of a common pool resource is managed through aggregated quota (TAC) and another is managed with private harvesting rights or IQ. In this environment, each group of harvesters must respond not only to the strategies of the other users facing their management incentives but also to the strategies of harvesters in the other management system. Note that the total number of harvesters (i.e., the sum of the two groups) is assumed fixed.

A. Single Rule Management

As a baseline for understanding the effects of having multiple concurrent management systems within a fishery, we characterize predictions for environments where all harvesters face the same management incentives. We set up a simple model where each individual harvester will maximize the total profit within a single fishing season, subject to the management system she/he faces. Two additional assumptions are made to simplify the model: first, we assume zero discount rate because this is a within-season model, following previous studies such as Boyce (2001) and Homans and Wilen (1997). Second, we abstract away from any stock effect on the harvesting function. With these assumptions, our maximization problem becomes a series of static optimizations throughout the fishing season, similar to the model described in Boyce (2001). (4) This simple model allows us to focus on the essential hypotheses that we later test in the experiment, namely the choice of effort level and harvest timing due to the difference in management systems.

Harvesters within an IQ management system plan their effort levels throughout a season of maximum length [T.sub.m] to solve the following problem:

(2) [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII]

With IQ, harvesters have a secure claim to a total seasonal quantity of harvest, [[bar.q].sub.i], and wish to time their landings so as to sell them at the highest price.

While the season length T is one of the choice variables, it can be shown that any season length shorter than the maximum allowed (i.e., [T.sup.*] < [T.sub.m]) is not part of the solution to this optimization problem (see Supporting Information, Appendix SI). This is to be expected, as in this environment the Nash equilibrium occurs when aggregate landings are at the same level in every period t: if landings are lower in one period than another, a harvester could receive a higher price by shifting her landings from a higher total landings period to the lower total landings period. (5) Thus, the IQ outcome is to stretch the season to its full length, landing the same amount in each period and attaining the efficient level of profit.

The interesting case is one where fishing is profitable such that the quota is a binding constraint on total effort. In this environment, and assuming identical individuals, the optimal harvest in each period is simply

(3) H([E.sup.*]) = ([N[bar.q]]/[T.sub.m]),

which in principle can be solved for optimal aggregated fishing effort, [E.sup.*].

The last case to consider is where even with the highest possible fishing effort, while still profitable, allocated quota is not exhausted before the season ends. This may occur, for example, if the maximum season length is significantly short compared to the amount of quota allocated. In this case, the optimal harvest volume and hence fishing effort can be solved from the following equation

(4) (1/N)[(a-2bH(E))(dH/dE)] + (N - 1/N) (a - 2bH (E)) x (H(E)/E) - [gamma] = 0.

One can easily verify that when N = 1 then Equation (4) becomes marginal revenue equal marginal cost, i.e., the sole owner solution. When N [right arrow] [infinity]. Equation (4) reduces to the open access solution where average revenue equals average cost.

In summary, IQ harvesters will always spread their fishing effort such that they operate during the entire season, i.e., [T.sup.*] = [T.sub.m]. If the season length is long enough compared to the quota allocated then each period's fishing effort will be as depicted in Equation (3). Otherwise the fishing effort will be as depicted in Equation (4), which is higher and can approach the derby fishing level depending on the group size (N).

When all harvesters are managed by a CP, the maximization problem is nearly identical to Equation (2) except for the first constraint, which becomes

(5) T [N.summation over (i=1)] ([e.sub.i]/E) H(E) [less than or equal to] [bar.Q],

reflecting that the total harvest among all harvesters cannot exceed the TAC ([bar.Q]), but without limiting the total harvest of any individual. The Nash equilibrium of this game is for each harvester to harvest as intensively as is economically feasible (until marginal revenue per unit effort reaches [gamma], or until [e.sub.max]) until the quota is exhausted, a fishing derby. Because landings are high in each period until the fishery closes, prices are low and the return to the resource is below the social optimum.

B. Multiple Rule Management

When some of the harvesters are managed by IQ and others are managed through a CP, each harvester must respond not only to the incentives of her own management system but also the behavior of harvesters in the other system. The incentives for the CP harvesters have not changed: lacking an ability to secure harvest by any means other than landing it, in equilibrium, CP harvesters engage in a full effort fishing derby until their quota is exhausted and they are shut down.

The IQ harvesters, however, can respond to the derby among CP harvesters. As shown in Equation (3) from the model, the IQ harvesters will adjust their landing in each period depending on the total season length available to them ([T.sub.m]), provided that [T.sub.m] is long enough to exhaust their quotas in that time. If [T.sub.m] is long enough--i.e., available quotas of the CP group are such that IQ harvesters can land less in every week remaining after CP shutdown than the CP did during its derby--then IQ harvesters will not fish during the derby and instead hold their quota to evenly distribute among the remaining periods once the CP is closed. This is driven by profit in each period being a decreasing function of the total effort of fishermen operating that period, and the result is relatively clean in our model because there is no discounting or stock effect on harvest productivity, and thus there is no opportunity cost of waiting until the derby fishing ends. Indeed, Scheld et al. (2012) found members of the RI Fluke Conservation Cooperative reduced their fluke harvest during derbies, choosing instead to land a large portion of their quota after the closure of the Summer subseason, when no one else was permitted to land and prices were 20% higher.

For the purpose of experimental design, however, we want to avoid the situation where the season length remaining after the CP harvesters exhaust their TAC is too short for IQ harvesters to harvest their entire quotas. In such case Equation (4) will prevail, and the difference in effort choice between CP and IQ harvesters becomes murky. On the basis of our model, we chose the parameter values so that CP harvesters will exhaust their aggregate TAC quickly enough so that the remaining periods will be sufficient for IQ harvesters to catch all their IQ at the effort level as shown in Equation (3).

III. EXPERIMENTAL DESIGN

To test these hypotheses about the effects of having multiple management systems operating within the same fishery, we designed a controlled economic experiment in which human subjects play the role of harvesters of a common pool resource being managed concurrently by CP and IQ. Each CP resource game represents a season, executed in 52 "weekly" effort decisions of the number of days in that week to fish, [e.sub.it] [member of]{0,1, ..., 7}. Each of 12 subjects has 3,000 pounds of quota associated with her, and each day of effort leads to landing 30 pounds of fish, [H.sub.t] ([E.sub.t]) = 30[E.sub.t]. Prices are determined by a downward sloping demand curve based on total weekly landings of all subjects (in Equation (1) above, a = 50 and b = 0.012).

Each experimental session included playing 15 seasons, of which three were randomly selected for payment at the end of the session. Each season was conducted in one of four treatments: In the CP treatment, all subjects are managed as a common pool, so any subject could fish as much as she/he wished, until the 36.000 pounds of aggregate quota was reached. In the IQ treatment, each subject was allocated 36,000 pounds of quota to harvest whenever she/he wished. In the Mixed treatment, six subjects were assigned to CP management (and would all be shut down once the total of CP subject landings reached 18,000 pounds) and six were assigned to IQ management, each assigned 3,000 pounds of IQ. In the Choice treatment, subjects were asked prior to the beginning of each season whether they wished to be in CP or IQ management: subjects choosing IQ received a 3,000 pound quota allocation, and subjects choosing CP had the 3,000 pounds associated with them placed in a common pool, which was shared among all subjects choosing group limit. (6)

Given these parameters, the Nash equilibrium strategies for players under each management system are presented in Table 1. In the CP treatment, all 12 subjects fish as intensively as possible to secure their share of the harvest, 7 days per week for the first 14 weeks, leaving them each only 2 days remaining, which is expended in Week 15. In the IQ treatment, the price externality is minimized by averaging 1.92 days per week for the entire season. In the Mixed treatment, the CP subjects race-to-fish as in the CP treatment. The IQ subjects would receive very low prices if they harvested during this derby, and thus they hold their quota until CP subjects are forced to reduce their effort, equating total landings--and thus price--across the remainder of the season.

[FIGURE 1 OMITTED]

In five sessions, there was a series of three CP-managed seasons, followed by three IQ-managed seasons, followed by four Mixed seasons (permuting assignments so each subject was in each group twice), and finally five Choice seasons. In the other four sessions, the first three seasons were IQ managed, followed by three CP-managed seasons in advance of the four Mixed and five Choice seasons. The last session of each group was of experienced subjects, who had participated in one of the earlier experiments (in either treatment ordering).

Figure 1 shows the interface of an IQ-managed subject in the Mixed treatment. The center of the screen shows that this subject is in the "Individual Limit" group, which has a total of 36,000 pounds of quota allocated, including information on how much others in the IQ group harvested the previous week and how much quota is remaining. To the left is the individual subject's You box (not shown to CP-managed subjects), which displays the subject's IQ allocation, previous week harvest, and quota remaining. To the right is the Other Group box, which shows the total quota allocation, previous week harvest, and quota remaining of the other group. The graph above the boxes tracks the subject's profit, and shows the quota remaining in each group. The Profit box in the lower right shows the detailed profit calculation, including individual landing times the prevailing price in the previous week.

Subjects make their decisions by clicking a radio button, corresponding to the "Number of days fishing per week" in the lower left-center of the screen. A novel feature of this environment is that it is quasi-continuous: weeks elapse every 4 s, capturing the effort level in the Days radio button panel, so subjects focus on when to change their effort level in response to their declining quota balance, the landings of others, and the amount of time remaining. This design has the advantage of allowing subjects a great deal of control over how to divide up the use of their 100 fishing days worth of quota, and an opportunity to respond to how others are using theirs. We are not aware of previous continuous CP resource experiments, but evidence from continuous time public goods experiments (e.g., Goren, Kurzban, and Rapoport 2003) suggests that continuous time may facilitate coordination on higher contributions to the public good, and shift outcomes slightly toward the social optimum.

[FIGURE 2 OMITTED]

IV. RESULTS

Figure 2 shows the predicted and average observed effort levels in each week in the last season of each treatment within a session, in the CP (top panel), IQ (middle panel), and Mixed (bottom panel) treatments. While the data do not precisely align with the predicted effort paths, key features of the predictions are present in the data that suggest the theoretical model is a useful organizing framework for the observed behavior. In the CP-managed fishery, the Nash equilibrium is for all subjects to choose the maximum possible effort level (7 in Weeks 1-14, 2 in Week 15), until the fishery is closed in Week 16. In the data, average effort levels are somewhat below the maximum effort level, and thus fishing continues for more weeks than predicted, but the level of effort is still quite high and in most sessions the season ends before Week 20. Under IQ management, Nash equilibrium predicts subjects will distribute their effort, and therefore their landings, equally throughout the season, so as to maximize the price received; this occurs at an average effort level just below 2. The data start slightly above 2, and gradually decrease toward 2 after Week 40, falling below 2 as some subjects exhaust their quotas due to higher initial effort levels. However, the overall impression is that effort is reasonably well distributed throughout the season, consistent with the prediction. In comparing the CP and IQ treatments, it is apparent that CP management induced a derby that led to much higher levels of effort and earlier closing, while IQ management reduced effort levels and lengthened the season considerably.

While the predictions in the CP and IQ treatments--where all subjects face the same management incentives--are consistent with the data, the Mixed treatment makes the more subtle prediction that IQ subjects will shift their strategy in response to the behavior of the CP group. The Nash equilibrium in the Mixed treatment for the CP-managed subjects is the same as in the CP treatment, a derby at full effort until the collective quota is exhausted. The IQ-managed subjects are predicted to avoid the low prices during the derby, and hold their quota until the CP fishery closes, when they then distribute their effort evenly across the balance of the season. The CP subjects' data are as consistent with the prediction as in the CP treatment, clearly reflecting a derby. While some IQ subjects do not delay their harvesting until the CP subjects are shut down, many do, and after the CP closure there is a notable increase in effort--to roughly the predicted level--among IQ subjects, which persists for the balance of the season. The next section establishes the statistical significance of the data patterns observed in Figure 2.

A. Single Management Systems

Table 2 presents the results of a two-level mixed-effects regression (incorporating season and session random effects) of total observed effort in all seasons of the CP and IQ treatments, as a function of treatment and experience indicators, and week-of-season time trend. The regression excludes observations in the first 3 weeks, the last 3 weeks, and the weeks of and following closure.

The Constant in the regression reflects effort levels in the CP-managed fishery, when it is open. Effort levels begin at 61.94, which is significantly below the predicted effort of 84 (7 times 12 subjects). However, the significantly positive coefficient on Week reflects that effort increases over time as the derby intensifies, possibly reflecting the collapse of initial efforts to coordinate on slower harvesting and higher prices.

The IQ variable reflects the difference between average effort in the CP and IQ treatments; it indicates average effort in the IQ treatment is 32.97, roughly half the average effort in the CP fishery (statistically different with p < [10.sup.-95]) and comparatively close to the predicted value of 23 (1.92 times 12 subjects). The significantly negative IQ x Week interaction, that is roughly twice the size of the Week variable, reflects a decrease in effort intensity over time in the IQ treatment. Efforts are nearly flat in the experienced session's IQ treatment, where they begin at 17.27 fishing days. A likelihood ratio test rejects the hypothesis of equal average efforts in the CP and IQ treatments with p < [10.sup.-56].

As a result of the higher effort in CP, the CP treatment seasons were shorter, as shown in Figure 3 (Wilcoxon rank-sum p = .002). The CP group, whether in the single-management or mixed-management treatment, exhausts its aggregated quota by around Week 19-21, while the IQ group nearly always extends its season to the maximum. With high weekly harvests, CP subjects' average profits were only $24,963, as compared with $71,852 in the IQ treatment (Wilcoxon signed-rank p < [10.sup.-18]). Thus, when all subjects are in the same management system, we do observe outcomes consistent with the general predictions that CP management leads to a low-profit derby and that IQ management leads to higher profits throughout a longer season.

[FIGURE 3 OMITTED]

B. Multiple Concurrent (Mixed) Management Systems

Table 3 presents the results of seemingly unrelated regressions of total effort among CP subjects (Equation (1)) and IQ subjects (Equation (2)), with bootstrapped standard errors with resampling by season to account for the nonindependence of observations from the same session. The independent variables are Week and Experience as in Table 2, with the addition of a CP_Closure variable that takes on a value of 1 if the CP-managed subjects have been closed in the week from which the observation is taken, and zero otherwise. In the CP equation, CP_Closure interactions offset the Constant, Week, and Experience variables so the predicted effort will be exactly zero when the fishery is closed; in the IQ equation, CP_Closure interactions show how IQ-managed subjects respond to the closure.

The only significant variable in the CP equation is Constant which, at 31.77, is lower than the predicted level of 42 (7 days times 6 subjects) and does not change with Week. In response to this nevertheless high and constant level of effort by the CP-managed subjects, the IQ-managed subjects put forth an average of only 10.96 days of effort prior to the CP closure; this is reduced to 4.36 among experienced subjects, close to the predicted effort level of zero (though a likelihood ratio test gives p = .015). Once the CP-managed subjects are closed, the IQ subjects' effort doubles, consistent with the Nash equilibrium, increasing to 22.40 among inexperienced subjects and 29.62 among those with experience. The insignificance of the Week terms in the IQ regression means total IQ subject effort remains constant for the balance of the season. Thus, while there is an overall tendency toward more moderate effort levels than theory predicts, the key dynamic predictions of the model are observed within the data.

As in the single management system treatments, the differences in effort levels induce statistically and economically significant differences in profitability. Subjects managed by CP in the Mixed treatment averaged $55,555 in profit, while those managed by IQ averaged $73,936 (Wilcoxon signed-rank p < [10.sup.-13]). Thus, having two management systems led to a harvest timing pattern consistent with that predicted by the model, with the resulting predicted differences in profit.

C. Management System Choice

As harvesters gain experience in, and observe outcomes in, the alternative management systems through successive seasons, they learn about their harvest and profit opportunities under each management system, and may develop preferences for one system or the other based on profitability or other properties. Having experienced outcomes under CP management alone, IQ management alone, and under each management system when competing against an equal number of subjects in the other, we measured these preferences by offering subjects the choice of the management system in which they would like to participate. The experiment parameters ensured Nash outcomes are always more profitable under IQ management regardless of the number of subjects who chose it.

On the basis of their prior experience, subjects chose to retain the 3,000 pounds of quota associated with them as an individual property right in 423 of the 540 instances (78.3% of the time). (7) This is different from random with p < [10.sup.-38]. Exactly half--54 of 108--of the subjects never chose CP management.

This is a strong preference for IQ management, but the persistence of a non-negligible number of subjects competing in a CP is puzzling. It also reflects an important element of real catch share programs, as in each of the cases discussed in Section I a significant number of harvesters chose not to join sectors when given the chance, often citing philosophical opposition to allocating property rights to the fishery. However, that is an unlikely motive among a selection of college student subjects. To understand the factors that contribute to a subject's choice to join a CP management group, Table 4 presents the results of a series of random effects logit models of management system choice, with the IQ alternative coded as one.

Economic theory would suggest that subjects' decisions should be determined by the subject's relative profitability under IQ and CP management. However, the InMaxProfitDif variable, which captures the difference in each subject's maximum profit in the IQ treatment and in the CP treatment, is not a significant factor in explaining variation in the choice of management in the Choice treatment. (8) Instead, the model identifies a series of lagged variables as significant determinants of the choice. A positive and significant coefficient on [NOthers.sub.t-1], the number of other subjects who chose CP in the last season, means subjects avoid larger CP in all models. The lagged dependent variable, [CPChosen.sub.t-1], is significant in Model 1 and indicates subjects who chose CP in the past are more likely to do it again, even controlling for the significant effect of past catch in the CP, [CPCatchi.sub.t-1]. However, measuring past catch instead by [CPOver3K.sub.t-1], a dummy variable for whether the subject was able to catch more than her equal 3,000 pound quota share last time she/he was in the CP group, swamps the significance of the lagged dependent variable. Adding another measure of the subject's aptitude under CP management, a significant negative coefficient on [CPMaxCatchMix.sub.i] indicates subjects who earn more in CP are more likely to choose it, independent of whether they are also better at IQ. Collectively, this model suggests that people who choose CP management are more likely to be "successful" in CP management, but they are defining success with respect to quantity landed rather than relative profitability.

V. DISCUSSION

To address an emerging generation of dynamic common pool resource (CPR) policy questions about the distribution of effort during a season induced by a fixed TAC which successfully manages interseason stock externalities we designed a novel dynamic CPR experimental environment in which subjects were playing the role of harvesters managed by either CP or IQ management. In the single management treatments, the data are broadly consistent with Nash equilibrium: there is a race-to-fish under CP management, and the high landings result in low prices and low profits; IQ-managed subjects distribute their landings throughout the season, receiving higher prices and earning higher profits. When subjects were split among IQ and CP management, the key features predicted by Nash equilibrium were again observed. First, the CP subjects still raced to fish, essentially unaffected by the presence of the IQ subjects. However, the IQ subjects limited their effort during the derby, and increased it significantly once the CP fishery was closed.

This strategic response of the IQ subjects to the presence of CP subjects is useful to observe in the laboratory for several reasons. First, that the Nash prediction is realized supports using our model as a framework for organizing the incentives of competing IQ and CP groups, and thus if the same incentives present themselves in the field, we would expect to observe similar comparative statics in actual harvester behavior. Second, from a management perspective, this change in effort carries practical importance: harvesters may have alternative fisheries they can pursue when they are not pursuing the sector-managed fishery, and thus the change to sector management can influence the degree of effort in other fisheries. That is, the IQ-managed harvesters who previously had to pursue the sector-managed fish during derbies can now find other species to pursue during those times. Scheld et al. (2012) found this is exactly what RI fluke sector vessels did, putting greater effort on skate and dogfish during the CP fluke derbies.

Our model and laboratory environment relies on perfect enforcement, ensuring individual vessels and groups of vessels stop fishing when the quota is met. Catch share management like that evaluated here has been adopted primarily in countries with functional governments, and resources for tracking and enforcing fishing activity at the vessel level (Melnychuk et al. 2012). Our motivating cases are large commercial fisheries in US territorial waters, where there is an extensive combination of on-board observers and dealer-based landings reporting to ensure catch is reported and counted against quota. While there are violations of fishing regulations, they do not alter the daily price externality (or interseason stock externality) to an economically significant extent. Thus, we expect our results to be broadly informative about the behavior and policy value in our applications of interest.

One of the major policy questions arising out of the initial experiences with catch share and sector programs is how they will expand, evolve, or fade away. The evolution of institutions has been a subject of recent experiments on the rule of law (e.g., Gurerk, Irlenbusch, and Rockenbach 2006). The RI fluke sector pilot grew through 2009-2011, as few people want to leave the sector, but others want to join; but the program was discontinued by the RI Department of Environmental Management in 2012. Similarly, the Chignik Coop attracted additional members, until courts declared its arrangement unconstitutional. The Cape Cod Hook and Cape Cod Fixed Gear sectors have seen their concept expanded throughout the Multispecies fishery, despite the resistance in the fishery to move to catch share management through the declines of the previous 30 years.

The Choice treatment allowed subjects to choose their management system based on what they had experienced and observed within each system. A strong preference for rights-based management emerged, suggesting that successful management groups will be able to attract more members, and to the extent that IQ management leads to success, IQ management will expand throughout the fishery. Viewed through this lens, sector-based catch share programs are not just a management tool, but a political tool for implementing effective management. They establish a process through which users can select the management methods that best suit their businesses, without needing to persuade regulators that they will be effective at anything other than keeping within the allocated catch share. If indeed they are effective, then the sector and its management will grow to be an important part of the fishery's management portfolio.

However, it is also often true that these sector programs were established to strike a political compromise between a subset of the industry interested in greater control over their harvesting than was afforded by current management, and by other participants who opposed individual quota allocation, or rights-based management altogether. In some cases, the sector groups were very large and held an overwhelming majority of the quota (e.g., Northeast Multispecies sectors left less than 5% of the quota for the CP; the Chignik Coop held over 80% of the quota), whereas in others, most of the quota remained in the CP (e.g., in Rhode Island, the sector held 11% of the quota). This gives rise to a potential future research question of how sectors with catch shares will adopt different management strategies depending on their relative dominance with respect to CP harvesters, and how the realized outcomes differ from those in the competitive CP fishery.

ABBREVIATIONS

CP: Common Pool (experimental treatment)

CPR: Common Pool Resource

IQ: Individual Quota (experimental treatment)

ITQ: Individual Transferable Quota

TAC: Total Allowable Catch

doi: 10.1111/ecin.12057

REFERENCES

Acheson, J., and R. Gardner. "Modeling Disaster: The Failure of Management of the New England Groundfish Industry." North American Journal of Fisheries Management, 31, 2011, 1005-18.

Agrawal, A. "Common Resources and Institutional Sustainability," in The Drama of the Commons, edited by E. Ostrom, T. Dietz, N. Dolsak, P. C. Stem, S. Stonich, and E. U. Weber. Washington, DC: National Academy Press, 2002.

Aoki, M., and Y. Hayami, eds. Communities and Markets in Economic Development. New York: Oxford University Press, 2001.

Arnason, R., K. Kelleher, and R. Willman. The Sunken Billions: The Economic Justification for Fisheries Reform. Washington, DC: The World Bank, 2009.

Baland, J.-M., and J.-P. Platteau. Halting Degradation of Natural Resources: Is There a Role for Rural Communities? New York: Oxford University Press, 1996.

Boyce, J. R. "Comment: Efficiency of ITQs in the Presence of Production Externalities." Marine Resource Economics, 15, 2001, 223-43.

Conrad, J. Resource Economics. 2nd ed. New York: Cambridge University Press, 2010.

Crosson, S. "Resistance to Alternative Management in Fisheries." Politics and the Life Sciences, 30, 2011, 31-42.

Cunningham, S., and T. Bostock, eds. Successful Fisheries Management: Issues, Case Studies and Perspectives. Delft, Holland: Eburon Academic Publishers, 2005.

Deacon, R., D. Parker, and C. Costello. "Improving Efficiency by Assigning Harvest Rights to Fishery Cooperatives: Evidence from the Chignik Salmon Co-op." Arizona Law Review, (50), 2008, 479-510.

Fell, H. "Ex-vessel Pricing and IFQs: A Strategic Approach." Marine Resource Economics, 24(4), 2009, 311-28.

Gardner, R., M. Moore, and J. Walker. "Governing a Groundwater Commons: A Strategic and Laboratory Analysis of Western Water Law." Economic Inquiry, 35, 1997, 218-34.

Georges Bank Cod Fixed Gear Sector. "Georges Bank Cod Fixed Gear Sector: A Final Environmental Assessment." 2010. Accessed September, 2012. http://nero.noaa.gov/sfd/sectordocs/EAs/GBCFG%20FY2010%20EA%20-%20FINAL.pdf.

Goren, H., R. Kurzban, and A. Rapoport. "Social Loafing versus Social Enhancement: Public Goods Provisioning in Real Time with Irrevocable Commitments." Organizational Behavior and Human Decision Processes, 90(2), 2003, 277-90.

Gurerk, O., B. Irlenbusch, and B. Rockenbach. "The Competitive Advantage of Sanctioning Institutions." Science, 312, 2006, 108-11.

Hanneson, R. "Sharing a Migrating Fish Stock." Marine Resource Economics, 28(1), 2013, 1-17.

Homans, F. R., and J. E. Wilen. "A Model of Regulated Open Access Resource Use." Journal of Environmental Economics and Management, 32, 1997, 1-21.

Hotvedt, B. 2011. "The Problem of Sharing a Common Stock: An Analysis of the Mackerel Conflict in the North East Atlantic." Masters Thesis, University of Tromso Norweigen College of Fisheries Science.

Kitts, A., E. Bing-Sawyer, J. Walden, C. Demarest, M. McPherson, P. Christman. S. Steinback, J. Olson, and P. Clay. 2010 Final Report on the Performance of the Northeast Multispecies (Groundfish) Fishery (May 2010-April 2011). Northeast Fisheries Science Center Reference Document 11-19, 2011.

Knapp, G. "The Chignik Salmon Cooperative," in Case Studies in Fisheries Self-Governance, edited by R. Townsend, R. Shotton, and H. Uchida. FAO Fisheries Technical Paper 504, 2008, 335-48.

Kuperan, K., and J. G. Sutinen. "Blue Water Crime: Legitimacy, Deterrence and Compliance in Fisheries." Law and Society Review, 32(2), 1998, 309-38.

Levhari, D., and L. Mirman. "The Great Fish War: An Example Using a Dynamic Cournot-Nash Solution." The Bell Journal of Economics, 11(1), 1980, 322-34.

Melnychuk, M., T. Essington, T. Branch, S. Heppell, O. Jensen, J. Link, S. Martell, A. Parma, J. Pope, and A. Smith. "Can Catch Share Fisheries Better Track Management Targets?" Fish and Fisheries, 13, 2012, 267-90.

National Research Council. The Drama of the Commons, edited by E. Ostrom, T. Dietz, N. Dolsak, P. C. Stern, S. Stonich, and E. U. Weber. Washington, DC: National Academy Press, 2002.

Ostrom, E. "A General Framework for Analyzing the Sustainability of Social-Ecological Systems." Science, 325, 2009, 419.

Pinto da Silva, P., and A. Kitts. "Collaborative Fisheries Management in the Northeast US: Emerging Initiatives and Future Directions." Marine Policy, 2006, 30, 832-41.

Scheld, A., C. Anderson, and H. Uchida. "The Economic Effects of Catch Share Management: The Rhode Island Fluke Sector Pilot Program." Marine Resource Economics, 27(3), 2012, 203-28.

Townsend, R., R. Shotton, and H. Uchida. Case Studies in Fisheries Self-Governance. FAO Fisheries Technical Paper 504, 2008.

Verani, A. "Community-Based Management of Atlantic Cod by the Georges Bank Hook Sector: Is It a Model Fishery?" Tulane Environmental Law Journal, 20, 2007, 359-80.

Walker, J., and R. Gardner. "Probabilistic Destruction of Common-Pool Resources: Experimental Evidence." The Economic Journal, 1992, 102, 1149-61.

Walker, J., R. Gardner, and E. Ostrom. "Rent Dissipation in a Limited Access Common Pool Resource: Experimental Evidence." Journal of Environmental Economics and Management, 1990, 19(3), 203-11.

Wilen, J. "Renewable Resource Economists and Policy: What Differences Have We Made?" Journal of Environmental Economics and Management, 2000, 39(3), 306-27.

Wilson, D. C., J. R. Nielsen, and P. Dengbol, eds. The Fisheries Co-management Experience: Accomplishments, Challenges and Prospects. Dordrecht, The Netherlands: Kluwer Academic Publishers, 2003.

SUPPORTING INFORMATION

Additional Supporting Information may be found in the online version of this article:

Appendix S1. Derivation of Nash Equilibrium Predictions

(1.) Summer flounder, or fluke, is a species that is jointly harvested with the Northeast Multispecies groundfish complex in southern New England, but is managed by the states even in federal waters.

(2.) This structure captures our motivating case of the RI fluke sector, but also characterizes the US/Canadian halibut and Virginia/North Carolina striped bass cases, and the Cape Cod sectors. The Chignik Coop served primarily to reduce effort costs, which is not well approximated by an individual quota sector.

(3.) Our choice of model setup that fisherman faces a downward sloping demand is based on our motivating case of the Rhode Island fluke fishery. See Scheld et al. (2012) for details.

(4.) Without the stock effect in the harvest or cost functions, there is no consequence of today's harvest on future productivity or cost, thus the temporal linkage with regard to harvest decisions does not exist. Furthermore, zero discounting means profit in each period has equal weight. Thus, the harvest will be the same in every period: if landings are lower in one period than another, a harvester could receive higher price by shifting her landings from a higher total landings (and lower price) period to the lower total landings period. Fell (2009) relaxes these assumptions, but must find Nash equilibrium numerically.

(5.) Any distribution of effort across individuals through time that leads to constant landings will satisfy the Nash equilibrium conditions. This model is formulated with symmetric agents, which makes it natural to think the symmetric equilibrium will be selected among the set of equilibrium strategies, but there is no equilibrium or refinement reason to expect so. (In fact, our experimental data show considerable asymmetry and effort level switching during a season, though with aggregate levels consistent with Nash equilibrium.)

(6.) For the parameters used in the experiment, it is strictly more profitable to be in the IQ managed sector than the CP managed sector in Nash equilibrium, regardless of the number of other subjects who choose IQ. The only exception is when all but one subject chooses IQ, in which case the lone CP subject will effectively have an IQ.

(7.) CP management was selected 117 times, though in one season, only one subject in the session chose CP, and thus effectively had an individual quota.

(8.) A wide range of other profit variables, including gross profit levels in the Mixed treatment, and the difference in average or maximum profit in the IQ and CP treatments were similarly insignificant.

CHRISTOPHER M. ANDERSON and HIROTSUGU UCHIDA *

* The authors are grateful to James Nugent for developing the experimental software, and to Mihoko Tegawa for research assistance. This work was supported by Rhode Island Sea Grant(#NA080AR4170691) and Rhode Island Agricultural Experiment Station (AES#5336).

Anderson: Associate Professor, School of Aquatic and Fishery Sciences, University of Washington, Seattle. WA 98195. Phone 1-(206)-543-1101, Fax 1-(206)-685-7471, E-mail cmand@uw.edu

Uchida: Assistant Professor, Department of Environmental & Natural Resource Economics, 205 Kingston Coastal Institute, University of Rhode Island, Kingston. RI 02881. Phone 1-(401)-874-2238, Fax 1-(401)-874-2156. E-mail uchida@uri.edu

TABLE 1
Nash Equilibrium Strategies for the CP,
IQ, and Mixed Treatments

                    Mean Nash
Treatment   Weeks     Days      Nash Profit

All CP
12 CP       1-14        7         $10,576
subjects     15         2
            16-52    Closed
All IQ
12 IQ       1-52      1.92        $75,077
Mixed
6 CP        1-14        7         $55,180
             15         2
            16-52    Closed
6 IQ        1-14        0         $82,619
             15       0.68
            16-52     2.68

TABLE 2
Mixed-Effects Model of Total Effort in the
Last Season of the CP and IQ Treatments

Variable                      Coefficient

Constant                         61.94
                                (51.34) ***
IQ                              -29.77
                                (20.69) ***
Week                              0.35
                                 (4.53) ***
Experience                        2.56
                                 (0.67)
IQ x Week                        -0.68
                                 (8.70) ***
IQ x Experience                 -14.91
                                 (3.30) ***
Week x Experience                 0.33
                                 (1.18)
Week x Experience x IQ            0.14
                                 (0.50)
LnL = -5549.65                  N = 1651
Wald [chi square] = 2323.45

Notes: Z statistics in parentheses. *** Significant at 1%.

TABLE 3
Seemingly Unrelated Regressions of Total Effort
in the Mixed Treatment

Variable                  CP            IQ

Constant                 31.77         10.96
                      (23.10) ***   (13.52) ***
CP_Closure              -31.77         11.44
                        (23.10)     (7.26) ***
Week                     -0.02         0.02
                        (0.21)        (0.37)
Experience               4.28          -6.60
                        (0.97)      (4.79) ***
CP_Closure x Week        0.02          -0.26
                        (0.21)      (3.85) ***
CP_Closure x             -4.28         7.22
  Experience            (0.97)        (1.22)
Week x Experience        -0.08         0.17
                        (0.21)        (-1.62)
Week x Experience x      0.08          -0.12
  CP_Closure            (0.21)        (0.48)
N = 1656
Pseudo-[R.sup.2]         0.951         0.198

Note: Z statistics are in parentheses. Significant
at *** 1% level.

TABLE 4
Random Effects Logit Model of Choice of IQ in Choice Treatment

                         Model 1      Model 2      Model 3

Constant                  10.73         2.33         4.96
                        (3.72) ***   (2.22) ***   (3.11) ***
InMaxProfitDif             0.07         0.07         0.09
                          (1.12)       (1.10)       (1.39)
[NOthers.sub.t-1]          0.35         0.38         0.38
                          (2.30)     (2.50) **    (2.48) **
[CPChosen.sub.t-1]        -8.07        0.243
                        (3.00) ***     (0.38)
[CPCatchi.sub.t-1]        -0.003
                        (3.85) ***
[CPOver3K.sub.t-1]                     -2.81        -2.77
                                     (3.71) ***   (4.70) ***
[CPMaxCatchMix.sub.i]                              -0.0006
                                                   (1.84) *
N = 432
InL                      -144.20      -150.19      -148.37
Wald [chi square]         33.81        29.66        32.28

Note: Z statistics are in parentheses. Significant at *** 1%, ** 5%,
and * 10% levels.